{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,9,5]],"date-time":"2026-09-05T04:45:55Z","timestamp":1788583555450,"version":"build-2803163510"},"reference-count":45,"publisher":"Springer Science and Business Media LLC","issue":"1","license":[{"start":{"date-parts":[[2026,7,3]],"date-time":"2026-07-03T00:00:00Z","timestamp":1783036800000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,7,3]],"date-time":"2026-07-03T00:00:00Z","timestamp":1783036800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"funder":[{"name":"Helmholtz Zentrum M\u00fcnchen - Deutsches Forschungszentrum f\u00fcr Gesundheit und Umwelt (GmbH)"}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Sci Rep"],"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>Large vision-language models (LVLMs) have shown impressive image-understanding capabilities across domains. However, their suitability for cytomorphological diagnostics remains unclear. Here, we systematically evaluated four state-of-the-art generalist LVLMs, GPT-4o, Gemini-2.0, Llama-3.2, and DeepSeek-VL2, and three biomedical LVLMs, LLaVA-Med, CONCH, and BiomedCLIP, across key cytomorphology benchmarks, including peripheral blood cell classification, morphology assessment, bone marrow cell classification, and cervical smear malignancy detection. Performance was assessed under zero-shot, few-shot, and fine-tuned settings. In zero-shot and few-shot evaluations, LVLMs performed poorly, often approaching random performance. In peripheral blood cell classification, GPT-4o achieved a zero-shot F1 score of only 0.22\u2009\u00b1\u20090.02 and a few-shot F1 score of 0.36\u2009\u00b1\u20090.03. Even after fine-tuning, GPT-4o was outperformed by a lightweight, dedicated hematology model. Beyond classification accuracy, we assessed interpretability and trustworthiness. Although LVLMs generated textual justifications, these often reflected textbook knowledge rather than the actual morphological features present in the cell images. Expert evaluation showed that 30% of explanations for misclassified cells were rated as poor or misleading. While LVLMs could segment cellular structures such as nuclei and granules, they failed to reliably identify the image regions relevant to their classification decisions. Our findings underscore three major limitations of current LVLMs in cytomorphology: (1) low diagnostic accuracy, (2) poor generalizability across domains, and (3) unreliable explainability. These results suggest that LVLMs require substantial improvement before they can be used for cell-type classification and morphology characterization in diagnostic settings. Purpose-built models remain the more effective and trustworthy choice.<\/jats:p>","DOI":"10.1038\/s41598-026-60372-6","type":"journal-article","created":{"date-parts":[[2026,7,3]],"date-time":"2026-07-03T07:05:07Z","timestamp":1783062307000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":2,"title":["Illusion of competence: vision\u2013language models provide confident but inaccurate explanations in cytological diagnostics"],"prefix":"10.1038","volume":"16","author":[{"given":"Ivan","family":"Kukuljan","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Muhammed Furkan","family":"Dasdelen","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Julia","family":"Sch\u00e4fer","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Michele","family":"Buck","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Katharina S.","family":"G\u00f6tze","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"ORCID":"https:\/\/orcid.org\/0000-0003-2154-4552","authenticated-orcid":false,"given":"Carsten","family":"Marr","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"297","published-online":{"date-parts":[[2026,7,3]]},"reference":[{"key":"60372_CR1","doi-asserted-by":"publisher","first-page":"498","DOI":"10.1056\/NEJMra043442","volume":"353","author":"BJ Bain","year":"2005","unstructured":"Bain, B. J. Diagnosis from the blood smear. N. Engl. J. Med. 353, 498\u2013507 (2005).","journal-title":"N. Engl. J. Med."},{"key":"60372_CR2","doi-asserted-by":"publisher","first-page":"193","DOI":"10.1016\/S0002-9378(16)40621-6","volume":"42","author":"GN Papanicolaou","year":"1941","unstructured":"Papanicolaou, G. N. & Traut, H. F. The diagnostic value of vaginal smears in carcinoma of the uterus. Am. J. Obstet. Gynecol. 42, 193\u2013206 (1941).","journal-title":"Am. J. Obstet. Gynecol."},{"key":"60372_CR3","doi-asserted-by":"publisher","first-page":"287","DOI":"10.1111\/ijlh.12327","volume":"37","author":"L Palmer","year":"2015","unstructured":"Palmer, L. et al. ICSH recommendations for the standardization of nomenclature and grading of peripheral blood cell morphological features. Int. J. Lab. Hematol. 37, 287\u2013303 (2015).","journal-title":"Int. J. Lab. Hematol."},{"key":"60372_CR4","volume-title":"Definitions, Criteria, and Explanatory Notes","author":"The Bethesda System for Reporting Cervical Cytology","year":"2015","unstructured":"The Bethesda System for Reporting Cervical Cytology. Definitions, Criteria, and Explanatory Notes (Springer International Publishing, Cham, 2015)."},{"key":"60372_CR5","doi-asserted-by":"publisher","first-page":"65","DOI":"10.1016\/j.jasc.2024.12.003","volume":"14","author":"S Satturwar","year":"2025","unstructured":"Satturwar, S. et al. American society of cytopathology\u2019s cytopathology workforce survey in the United States. J. Am. Soc. Cytopathol. 14, 65\u201377 (2025).","journal-title":"J. Am. Soc. Cytopathol."},{"key":"60372_CR6","doi-asserted-by":"publisher","first-page":"4369","DOI":"10.1038\/s41467-024-48705-3","volume":"15","author":"J Wang","year":"2024","unstructured":"Wang, J. et al. Artificial intelligence enables precision diagnosis of cervical cytology grades and cervical cancer. Nat. Commun. 15, 4369 (2024).","journal-title":"Nat. Commun."},{"key":"60372_CR7","doi-asserted-by":"publisher","first-page":"19535","DOI":"10.1038\/s41598-024-70670-6","volume":"14","author":"Y Kurita","year":"2024","unstructured":"Kurita, Y. et al. Enhancing cervical cancer cytology screening via artificial intelligence innovation. Sci. Rep. 14, 19535 (2024).","journal-title":"Sci. Rep."},{"key":"60372_CR8","doi-asserted-by":"crossref","unstructured":"Chowdary, G. J., G, S., M, P. & Yogarajah, P. Nucleus segmentation and classification using residual SE-UNet and feature concatenation approach incervical cytopathology cell images. Technol. Cancer Res. Treat. 22, 15330338221134833 (2023).","DOI":"10.1177\/15330338221134833"},{"key":"60372_CR9","doi-asserted-by":"publisher","DOI":"10.3390\/diagnostics13071363","volume":"13","author":"M Kalbhor","year":"2023","unstructured":"Kalbhor, M., Shinde, S., Popescu, D. E. & Hemanth, D. J. Hybridization of deep learning pre-trained models with machine learning classifiers and fuzzy min\u2013max neural network for cervical cancer diagnosis. Diagnostics 13, 1363 (2023).","journal-title":"Diagnostics"},{"key":"60372_CR10","doi-asserted-by":"publisher","first-page":"1917","DOI":"10.1182\/blood.2020010568","volume":"138","author":"C Matek","year":"2021","unstructured":"Matek, C., Krappe, S., M\u00fcnzenmayer, C., Haferlach, T. & Marr, C. Highly accurate differentiation of bone marrow cell morphologies using deep neural networks on a large image data set. Blood 138, 1917\u20131927 (2021).","journal-title":"Blood"},{"key":"60372_CR11","doi-asserted-by":"publisher","first-page":"45","DOI":"10.1038\/s43856-022-00107-6","volume":"2","author":"RM Tayebi","year":"2022","unstructured":"Tayebi, R. M. et al. Automated bone marrow cytology using deep learning to generate a histogram of cell types. Commun. Med. 2, 45 (2022).","journal-title":"Commun. Med."},{"key":"60372_CR12","doi-asserted-by":"publisher","first-page":"538","DOI":"10.1038\/s42256-019-0101-9","volume":"1","author":"C Matek","year":"2019","unstructured":"Matek, C., Schwarz, S., Spiekermann, K. & Marr, C. Human-level recognition of blast cells in acute myeloid leukaemia with convolutional neural networks. Nat. Mach. Intell. 1, 538\u2013544 (2019).","journal-title":"Nat. Mach. Intell."},{"key":"60372_CR13","doi-asserted-by":"publisher","DOI":"10.1016\/j.cmpb.2019.105020","volume":"180","author":"A Acevedo","year":"2019","unstructured":"Acevedo, A., Alf\u00e9rez, S., Merino, A., Puigv\u00ed, L. & Rodellar, J. Recognition of peripheral blood cell images using convolutional neural networks. Comput. Methods Programs Biomed. 180, 105020 (2019).","journal-title":"Comput. Methods Programs Biomed."},{"key":"60372_CR14","doi-asserted-by":"publisher","first-page":"2562","DOI":"10.1038\/s41598-023-29160-4","volume":"13","author":"P Manescu","year":"2023","unstructured":"Manescu, P. et al. Detection of acute promyelocytic leukemia in peripheral blood and bone marrow with annotation-free deep learning. Sci. Rep. 13, 2562 (2023).","journal-title":"Sci. Rep."},{"key":"60372_CR15","doi-asserted-by":"publisher","first-page":"111","DOI":"10.1038\/s41375-021-01408-w","volume":"36","author":"J-N Eckardt","year":"2022","unstructured":"Eckardt, J.-N. et al. Deep learning detects acute myeloid leukemia and predicts NPM1 mutation status from bone marrow smears. Leukemia 36, 111\u2013118 (2022).","journal-title":"Leukemia"},{"key":"60372_CR16","doi-asserted-by":"publisher","first-page":"201","DOI":"10.1186\/s12885-022-09307-8","volume":"22","author":"J-N Eckardt","year":"2022","unstructured":"Eckardt, J.-N. et al. Deep learning identifies Acute Promyelocytic Leukemia in bone marrow smears. BMC Cancer 22, 201 (2022).","journal-title":"BMC Cancer"},{"key":"60372_CR17","doi-asserted-by":"publisher","first-page":"38","DOI":"10.1038\/s41698-021-00179-y","volume":"5","author":"J-W Sidhom","year":"2021","unstructured":"Sidhom, J.-W. et al. Deep learning for diagnosis of acute promyelocytic leukemia via recognition of genomically imprinted morphologic features. Npj Precis. Oncol. 5, 38 (2021).","journal-title":"Npj Precis. Oncol."},{"key":"60372_CR18","doi-asserted-by":"publisher","first-page":"1318","DOI":"10.1038\/s41375-026-02934-1","volume":"40","author":"MF Dasdelen","year":"2026","unstructured":"Dasdelen, M. F. et al. AI-based hematological malignancy prediction from peripheral blood smears in a large diagnostic laboratory cohort. Leukemia 40, 1318\u20131322 (2026).","journal-title":"Leukemia"},{"key":"60372_CR19","doi-asserted-by":"publisher","DOI":"10.1371\/journal.pdig.0000187","volume":"2","author":"M Hehr","year":"2023","unstructured":"Hehr, M. et al. Explainable AI identifies diagnostic cells of genetic AML subtypes. PLoS Digit. Health 2, e0000187 (2023).","journal-title":"PLoS Digit. Health"},{"key":"60372_CR20","doi-asserted-by":"publisher","first-page":"520","DOI":"10.1007\/978-3-031-72390-2_49","volume-title":"Medical Image Computing and Computer Assisted Intervention\u2014MICCAI 2024","author":"V Koch","year":"2024","unstructured":"Koch, V. et al. DinoBloom: A Foundation Model for Generalizable Cell Embeddings in Hematology. In Medical Image Computing and Computer Assisted Intervention\u2014MICCAI 2024 (eds Linguraru, M. G. et al.) 520\u2013530 (Springer Nature Switzerland, Cham, 2024). https:\/\/doi.org\/10.1007\/978-3-031-72390-2_49."},{"key":"60372_CR21","doi-asserted-by":"publisher","first-page":"1930","DOI":"10.1038\/s41591-023-02448-8","volume":"29","author":"AJ Thirunavukarasu","year":"2023","unstructured":"Thirunavukarasu, A. J. et al. Large language models in medicine. Nat. Med. 29, 1930\u20131940 (2023).","journal-title":"Nat. Med."},{"key":"60372_CR22","doi-asserted-by":"publisher","DOI":"10.1038\/s43856-023-00370-1","volume":"3","author":"J Clusmann","year":"2023","unstructured":"Clusmann, J. et al. The future landscape of large language models in medicine. Commun. Med. 3, 141 (2023).","journal-title":"Commun. Med."},{"key":"60372_CR23","doi-asserted-by":"publisher","DOI":"10.2196\/59069","volume":"27","author":"K Zhang","year":"2025","unstructured":"Zhang, K. et al. Revolutionizing health care: The transformative impact of large language models in medicine. J. Med. Internet Res. 27, e59069 (2025).","journal-title":"J. Med. Internet Res."},{"key":"60372_CR24","doi-asserted-by":"publisher","DOI":"10.1016\/j.isci.2024.109713","author":"X Meng","year":"2024","unstructured":"Meng, X. et al. The application of large language models in medicine: A scoping review. iScience https:\/\/doi.org\/10.1016\/j.isci.2024.109713 (2024).","journal-title":"iScience"},{"key":"60372_CR25","doi-asserted-by":"publisher","first-page":"4886","DOI":"10.1038\/s41467-025-60014-x","volume":"16","author":"M Tran","year":"2025","unstructured":"Tran, M. et al. Generating dermatopathology reports from gigapixel whole slide images with HistoGPT. Nat. Commun. 16, 4886 (2025).","journal-title":"Nat. Commun."},{"key":"60372_CR26","doi-asserted-by":"publisher","unstructured":"Jiang, Y. et al. Evaluating General Vision-Language Models for Clinical Medicine. 2024.04.12.24305744 Preprint at https:\/\/doi.org\/10.1101\/2024.04.12.24305744 (2024).","DOI":"10.1101\/2024.04.12.24305744"},{"key":"60372_CR27","unstructured":"Royer, C., Menze, B. & Sekuboyina, A. MultiMedEval: A Benchmark and a Toolkit for Evaluating Medical Vision-Language Models. In: Proceedings of The 7nd International Conference on Medical Imaging with Deep Learning 1310\u20131327 (PMLR, 2024)."},{"key":"60372_CR28","doi-asserted-by":"publisher","DOI":"10.3389\/frai.2024.1430984","author":"I Hartsock","year":"2024","unstructured":"Hartsock, I. & Rasool, G. Vision-language models for medical report generation and visual question answering: A review. Front. Artif. Intell. https:\/\/doi.org\/10.3389\/frai.2024.1430984 (2024).","journal-title":"Front. Artif. Intell."},{"key":"60372_CR29","doi-asserted-by":"publisher","unstructured":"Jeong, D. P., Garg, S., Lipton, Z. C. & Oberst, M. Medical adaptation of large language and vision-language models: Are we making progress? Preprint at https:\/\/doi.org\/10.48550\/arXiv.2411.04118 (2024).","DOI":"10.48550\/arXiv.2411.04118"},{"key":"60372_CR30","doi-asserted-by":"publisher","first-page":"7866","DOI":"10.1038\/s41467-025-62385-7","volume":"16","author":"C Wu","year":"2025","unstructured":"Wu, C. et al. Towards generalist foundation model for radiology by leveraging web-scale 2D&3D medical data. Nat. Commun. 16, 7866 (2025).","journal-title":"Nat. Commun."},{"key":"60372_CR31","doi-asserted-by":"publisher","DOI":"10.1056\/AIoa2300138","volume":"1","author":"T Tu","year":"2024","unstructured":"Tu, T. et al. Towards generalist biomedical AI. NEJM AI 1, AIoa2300138 (2024).","journal-title":"NEJM AI"},{"key":"60372_CR32","doi-asserted-by":"publisher","first-page":"4344","DOI":"10.1109\/TMI.2024.3419697","volume":"43","author":"D Cai","year":"2024","unstructured":"Cai, D. et al. Hicervix: An extensive hierarchical dataset and benchmark for cervical cytology classification. IEEE Trans. Med. Imaging 43, 4344\u20134355 (2024).","journal-title":"IEEE Trans. Med. Imaging"},{"key":"60372_CR33","doi-asserted-by":"publisher","DOI":"10.1016\/j.dib.2020.105474","volume":"30","author":"A Acevedo","year":"2020","unstructured":"Acevedo, A. et al. A dataset of microscopic peripheral blood cell images for development of automatic recognition systems. Data Brief 30, 105474 (2020).","journal-title":"Data Brief"},{"key":"60372_CR34","doi-asserted-by":"crossref","unstructured":"Tsutsui, S., Pang, W. & Wen, B. WBCAtt: A white blood cell dataset annotated with detailed morphological attributes. In Proceedings of the 37th Conference on Neural Information Processing Systems (NeurIPS 2023), 50796\u201350824 (2023)","DOI":"10.52202\/075280-2210"},{"key":"60372_CR35","doi-asserted-by":"publisher","first-page":"1773","DOI":"10.1038\/s41597-025-06223-x","volume":"12","author":"S Shetab Boushehri","year":"2025","unstructured":"Shetab Boushehri, S. et al. A large expert-annotated single-cell peripheral blood dataset for hematological disease diagnostics. Sci. Data 12, 1773 (2025).","journal-title":"Sci. Data"},{"key":"60372_CR36","unstructured":"Hello GPT-4o. OpenAI https:\/\/openai.com\/index\/hello-gpt-4o\/."},{"key":"60372_CR37","doi-asserted-by":"publisher","unstructured":"Team, G. et al. Gemini: A Family of Highly Capable Multimodal Models. Preprint at https:\/\/doi.org\/10.48550\/arXiv.2312.11805 (2025).","DOI":"10.48550\/arXiv.2312.11805"},{"key":"60372_CR38","doi-asserted-by":"publisher","unstructured":"Grattafiori, A. et al. The Llama 3 Herd of Models. Preprint at https:\/\/doi.org\/10.48550\/arXiv.2407.21783 (2024).","DOI":"10.48550\/arXiv.2407.21783"},{"key":"60372_CR39","doi-asserted-by":"publisher","unstructured":"Wu, Z. et al. DeepSeek-VL2: Mixture-of-Experts Vision-Language Models for Advanced Multimodal Understanding. Preprint at https:\/\/doi.org\/10.48550\/arXiv.2412.10302 (2024).","DOI":"10.48550\/arXiv.2412.10302"},{"key":"60372_CR40","doi-asserted-by":"publisher","unstructured":"Li, C. et al. LLaVA-Med: Training a Large Language-and-Vision Assistant for Biomedicine in One Day. Preprint at https:\/\/doi.org\/10.48550\/arXiv.2306.00890 (2023).","DOI":"10.48550\/arXiv.2306.00890"},{"key":"60372_CR41","doi-asserted-by":"publisher","first-page":"863","DOI":"10.1038\/s41591-024-02856-4","volume":"30","author":"MY Lu","year":"2024","unstructured":"Lu, M. Y. et al. A visual-language foundation model for computational pathology. Nat. Med. 30, 863\u2013874 (2024).","journal-title":"Nat. Med."},{"key":"60372_CR42","doi-asserted-by":"publisher","DOI":"10.1056\/AIoa2400640","volume":"2","author":"S Zhang","year":"2025","unstructured":"Zhang, S. et al. A multimodal biomedical foundation model trained from fifteen million image\u2013text pairs. NEJM AI 2, AIoa2400640 (2025).","journal-title":"NEJM AI"},{"key":"60372_CR43","unstructured":"Fine-tuning now available for GPT-4o. OpenAI https:\/\/openai.com\/index\/gpt-4o-fine-tuning\/."},{"key":"60372_CR44","volume-title":"Mosby\u2019s Manual of Diagnostic and Laboratory Tests\u2014E-Book","author":"KD Pagana","year":"2009","unstructured":"Pagana, K. D. & Pagana, T. J. Mosby\u2019s Manual of Diagnostic and Laboratory Tests\u2014E-Book (Elsevier Health Sciences, 2009)."},{"key":"60372_CR45","doi-asserted-by":"publisher","first-page":"7995","DOI":"10.1038\/s41598-021-86995-5","volume":"11","author":"JPE Schouten","year":"2021","unstructured":"Schouten, J. P. E. et al. Tens of images can suffice to train neural networks for malignant leukocyte detection. Sci. Rep. 11, 7995 (2021).","journal-title":"Sci. Rep."}],"container-title":["Scientific Reports"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/www.nature.com\/articles\/s41598-026-60372-6","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41598-026-60372-6.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/www.nature.com\/articles\/s41598-026-60372-6.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,7,3]],"date-time":"2026-07-03T07:25:18Z","timestamp":1783063518000},"score":1,"resource":{"primary":{"URL":"https:\/\/www.nature.com\/articles\/s41598-026-60372-6"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,7,3]]},"references-count":45,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2026,12]]}},"alternative-id":["60372"],"URL":"https:\/\/doi.org\/10.1038\/s41598-026-60372-6","relation":{},"ISSN":["2045-2322"],"issn-type":[{"value":"2045-2322","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,7,3]]},"assertion":[{"value":"27 June 2025","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"26 June 2026","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"3 July 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}},{"order":1,"name":"Ethics","group":{"name":"EthicsHeading","label":"Declarations"}},{"value":"The authors declare no competing interests.","order":2,"name":"Ethics","group":{"name":"EthicsHeading","label":"Competing interests"}},{"value":"This study did not require ethical approval.","order":3,"name":"Ethics","group":{"name":"EthicsHeading","label":"Ethics approval"}}],"article-number":"20526"}}